AcciMap causal analysis of Chinese chemical industry accidents unraveled by graph neural networks

As global industrialization accelerates, chemical accidents present significant threats to production safety due to their destructive coupling of multiple hazards. To investigate the key causes contributing to chemical accidents, this study collected 103 chemical accident cases in China from 2015 to 2023. Each case was analyzed using the AcciMap methodology. A graph neural network was subsequently trained on the 103 AcciMap diagrams to generate causal importance scores for the causes contributing to chemical accidents. The results indicate that the complexity of production processes and environments in chemical plants leads to diverse accident development pathways. Therefore, preventive measures should be designed with tailored safety management plans specific to the characteristics of each chemical plant. Among the secondary causes, failure in material control exhibited the highest causal importance score of 8.56, indicating that unsafe condition of materials is the primary direct cause of chemical accidents. Among the primary causes, relevant enterprise management had the highest causal importance score of 32.14, significantly surpassing personnel activities and equipment, which are often regarded as direct causes. This highlights the advantage of the AcciMap approach in systemic safety analysis. This study proposes a model for applying big data techniques in the domain of system safety.